摘要
为了能够简单并且有效地识别红外图像序列中的运动目标,提出了一种新颖的基于贝叶斯概率计算的目标识别方法。在初始帧中利用相关算法实现对目标的初始定位,分析当前目标识别属性,建立判别函数。计算当前帧中连通区域的概率判断其为目标类或者背景类。对当前帧中的目标定位后,更新模式向量,用于下一帧中该类目标的识别。实验结果表明,该方法根据目标的识别属性,通过概率计算能够快速有效地识别运动目标,计算量小,所涉及的算法适于嵌入式系统实现,具有较好的鲁棒性。
To recognize moving targets simply and efficiently in infrared image sequences,a novel target recognition method based on Bayesian probability theory is proposed.The target initial position in the original image frame is obtained using the correlation matching algorithm.According to the recognition properties of the target,the decision functions are established.The classification of the connected regions in the current frame is determined by calculating Bayesian probability.The pattern vector is updated after the new target is located in the current frame for the target recognition in the next frame,and the purpose of moving target recognition is achieved.Experimental results show that:according to the recognized properties and probability calculation,the method can recognize the moving target fast and effectively with low computational complexity,and also the algorithm used is suitable for embedded system complementation and has strong robusticity.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2013年第1期76-80,共5页
Journal of Nanjing University of Science and Technology
关键词
运动目标
红外图像序列
目标识别
贝叶斯概率
鲁棒性
moving targets
infrared image sequences
target recognition
Bayesian probability
robusticity
